Benchmarking Edge AI Platforms: Performance Analysis of NVIDIA Jetson and Raspberry Pi 5 with Coral TPU

The rapid development of edge AI applications has led to the demand for high-performance, power-efficient embedded computing platforms. This paper evaluates the computational efficiency, power consumption, and thermal characteristics of the NVIDIA Jetson Orin NX, Jetson Nano, and Raspberry Pi 5 with Coral TPU across various AI workloads. Performance metrics, including inference speed, CPU and GPU utilization, and thermal stability, were benchmarked using a YOLOv5 object detection model. The results indicate that the Jetson Orin NX achieves the highest frame rates, nearly doubling the performance of the Raspberry Pi 5 with Coral TPU (41.8 FPS vs. 21.5 FPS) and surpassing the Jetson Nano by more than 50 percent. Power efficiency was another critical factor, with the Jetson Nano consuming the least power at approximately 7W, while the Raspberry Pi 5 with Coral TPU consumed 8.3W and the Jetson Orin NX reached 10.6W under load. Thermal analysis revealed that the Raspberry Pi 5 operates at significantly higher temperatures, reaching up to 80°C without active cooling, whereas the Jetson Nano and Jetson Orin NX maintained stable temperatures at 42°C and 45°C, respectively. These findings underscore the importance of active cooling solutions and optimized workload distribution for edge AI devices. The comparative assessment provided in this study helps guide hardware selection for specific edge AI applications, balancing trade-offs between power efficiency, thermal performance, and computational speed.